On the exceptional set of Lagrange’s equation
with three prime and
onealmost-prime variables
par DOYCHIN TOLEV
RÉSUMÉ. Nous considérons une version affaiblie de la conjecture
sur la représentation des entiers comme somme de quatre carrés de nombres premiers.
ABSTRACT. We consider an approximation to the
popular
con- jecture about representations of integers as sums of four squares of prime numbers.1. Introduction and statement of the result
The famous theorem of
Lagrange
states that everynon-negative integer
n can be
represented
aswhere ~i,...,.r4 are
integers.
There is aconjecture,
which asserts thatevery
sufficiently large integer
n, such that n - 4(mod 24~,
can be repre-sented in the form
(1.1)
withprime
variables ~1, .. , x4. Thisconjecture
has not been
proved
sofar,
but there are variousapproximations
to itestablished.
We have to mention first that in 1938 Hua
[9] proved
thesolvability
ofthe
corresponding equation
with fiveprime
variables. In 1976 Greaves[3]
and later Shields
[19]
and Plaksin[18] proved
thesolvability
of(1.1)
withtwo
prime
and twointeger
variables(in [18]
and[19]
anasymptotic
formulafor the number of solutions was
found).
In 1994 Brfdern and
Fouvry [2]
considered(1.1)
withalmost-prime
vari-ables and
proved
that if rt islarge enough
and satisfies n - 4(mod 24)
then(l.l)
has solutions inintegers
oftype P34.
Here and later we denoteby P r
anyinteger
with no more than rprime factors,
countedaccording
tomultiplicity. Recently
Heath-Brown and the author[8] proved that,
underthe same conditions on rt, the
equation (l.l)
has solutions in oneprime
andthree
P101 - almost-prime
variables and also in fourP25 - almost-prime
variables. These results were
sharpened slightly by
the author~21~,
wholB/Ianuscrit reçu le 7janvier 2004.
established the
solvability
of(1.1)
in oneprime
and threeP80 -
almost-primes and, respectively,
in fourP21 - almost-primes.
There are several papers,
published during
the last years, devoted tothe
study
of theexceptional
set of theequation (1.1)
withprime
variables.Suppose
that Y is alarge
real number and denoteby El (Y)
the numberof
positive integers
n Ysatisfying
n - 4(mod 24)
and which cannotbe
represented
in the form(1.1)
withprime
variables ~1, ... , ~4. In 2000J.Liu and M.-C. Liu
[16] proved
thaty13/15+é,
where e > 0 isarbitrarily
small. This result wasimproved considerably by Wooley [22],
who established that
El(Y)
«y13/30+é. Recently
L. Liu[15]
established thatEl (Y)
«y2/5+é.
In the paper
[22] Wooley
obtained otllerinteresting results, concerning
the
equation (1.1).
We shall state one of them. Denoteby R(n)
the numberof solutions of
(1.1)
in threeprime
and oneinteger
variables. It isexpected
that
R(n)
can beapproximated by
thewhere
C~(n)
is thecorresponding singular
series(see [22]
for thedefinition).
Wooley proved
that the set ofintegers
n, for whichR(n)
fails to be closeto the
expected value,
isremarkably
thin. Moreprecisely,
let0(t)
be anymonotonically increasing
andtending
toinfinity
function of thepositive
variable
t,
such that «(logt)B
for some constant B > 0. Let Y be alarge
real number and denoteby
the number ofpositive integers
Y such that
Theorem 1.2 of
[22]
asserts thatWe note that if the
integer n
satisfiesTherefore,
ifE2(Y)
denotes the number ofpositive integers n Y,
satis-fying (1.2)
and which cannot berepresented
in the form(1.1)
with threeprime
and oneinteger variables,
thenfor any é > 0.
The purpose of the
present
paper is to obtain an estimate of almost thesame
strength
as(1.3)
for theexceptional
set of theequation (1.1)
withthree
prime
and onealmost-prime
variables. We shall prove thefollowing
Theorem. Let Y be a
large
real number and denoteby E(Y)
the numberof positive integers
n Ysatisfying
n - 4(mod 24)
and which cannot berepresented
in theform
where P1,P2,P3 are
primes
and x =P11.
Then we haveAs one may
expect,
theproof
of this result istechnically
morecompli-
cated than the
proof
of Theorem 1.2 of[22].
We use a combination of the circle method and the sieve methods.In the circle method
part
weapply
theapproach of Wooley [22], adapted
for our needs. On the set of minor arcs we
apply
the method of Klooster-man, introduced in the classical paper
[14].
Thistechnique
was,actually,
applied
alsoby Wooley
in the estimation of the sumsTi
andT2
in section 3 of[22].
In hisanalysis, however, only Ramanujan’s
sums appear, whiles inour situation we have to deal with much more
complicated
sums, definedby (8.19). Fortunately,
these sums differ veryslightly
from sums consideredby
Brfdern andFouvry [2],
so we can, infact,
borrow their result for our needs.The sieve method
part
is rather standard. Weapply
aweighted
sievewith
weights
of Richert’s type andproceed
as inchapter
9 of Halberstamand Richert’s book
[4].
In many
places
we omit the calculations becausethey
are similar to thosein other books or papers, or because
they
are standard andstraightforward.
We note that one can obtain
slightly stronger
result(with
smaller power in(1.5)
and with variable xhaving
fewerprime factors) by
means of moreelaborate
computational
work.Acknowledgement.
The main part of this paper was writtenduring
thevisit of the author to the Institute of Mathematics of the
University
ofTsukuba. The author would like to thank the
Japan Society
of Promotionof Science for the financial support and to the staff of the Institute for the excellent
working
conditions. The author isespecially grateful
to ProfessorHiroshi Mikawa for the
interesting
discussions and valuable comments.The author would like to thank also Professor Trevor
Wooley
for inform-ing
about his paper[22]
andproviding
with themanuscript.
2. Notations and some definitions
Throughout
the paper we use standard number-theoretic notations. Asusual, p(n)
denotes the M6biusfunction,
is the Eulerfunction,
is the number of
prime
divisors of n, countedaccording
tomultiplicity,
T(n)
is the number ofpositive
divisors of n. Thegreatest
common divisorand, respectively,
the least commonmultiple
of theintegers
MI, M2 aredenoted
by (m1, m2)
and[m1, m2]. However,
if u and v are real numbers then(u, v)
means the interval withendpoints u
and v. Themeaning
isalways
clear from the context. We use boldstyle
letters to denote four-dimensional vectors. The letter p is reserved for
prime
numbers.If p
> 2then -
stands for theLegendre symbol.
To denote summation over thepositive integers
n Z we writeFurthermore, ~x(~,), respectively,
¿x(q)*
means that x runs over acomplete, respectively,
reducedsystem
of residues modulo q.By ~a~
we denote theinteger part
of the real number a,e(~) - e"?"
andeq(a)
=e(a/q).
We assume that E > 0 is an
arbitrarily
smallpositive
number and Ais an
arbitrarily large number; they
can take different values in different formulas. Unless it is notspecified explicitly,
the constants in the 0 -terms and « -
symbols depend
on E and A. Forpositive
U and V we write!7 x V as an abbreviation of U « V « U.
Let N be a
suffciently large
real number. We definewhere E > 1 is a
large
constant, which we shallspecify
later.To
apply
the sieve method we need information about the number of solutions of(1.4)
inintegers
xlying
in arithmeticalprogressions
and inprimes
pl, p2, p3. For technical reasons we attachlogarithmic weights
tothe
primes
and a smoothweight
to the variable x. Moreprecisely,
weconsider the function
and let
For any
integer n
E(N/2, N]
and for anysquarefree integer k,
such that(k, 6)
=1,
we defineWe expect that this
quantity
can beapproximated,
at least on average,by
the
expression
which arises as a mean term when we
apply formally
the circle method.The
quantity K(n)
from theright-hand
side of(2.6)
comes from thesingular integral
and is definedby
Having
in mind(2.3)
we see thatFurthermore, 6 (n, ~;)
comes from thesingular
series and is definedby
where
We shall consider
C‘~(7t, k)
in detail in the next section.3. Some
properties
of the sum6(n, k)
It is not difficult to see that the function
f (q, n,1~),
definedby (2.10),
ismultiplicative
withrespect
to q. We shallcompute
it for q =p~.
From this
point
onwards we assume that theintegers
n and ksatisfy
Then we have
Furthermore, if p
> 2 is aprime,
thenand
where the
quantities hj
are definedby
The
proof
of formulas(3.2) - (3.8)
is standard and usesonly
the basicproperties
of the Gauss sums(see
Hua[10], chapter 7,
forexample).
Weleave the verification to the reader.
From
(3.2) - (3.8)
weeasily get
This estimate
implies
that the series(2.9)
isabsolutely convergent.
Weapply
Euler’sidentity
and we use(3.1) - (3.5)
to obtainFrom this formula and
(3.4),
after somerearrangements,
weget
where
and where
From
(3.1), (3.5) - (3.8), (3.11)
and(3.13)
we obtain the estimatesand
where the constant in the 0 -term in the last formula is absolute. We leave the easy verification of formulas
(3.14) - (3.17)
to the reader.Let us note that from
(3.10), (3.12)
and(3.16)
it followswhich we, of course,
expect, having
in mind the definition(2.5)
ofI(n, k)
and the conditions
(3.1).
4. Proof of the Theorem
A central role in the
proof
of the Theoremplays
thefollowing Proposi- tion,
which asserts that the difference between thequantities I(n, k)
andE(n, ~ ),
definedby (2.5)
and(2.6),
is small on average withrespect
to nand k.
Proposition. Suppose
that the set F consistsof integers
n E(N/2, N],
satisfying
the congruence n - 4(mod 24),
and denoteby
F thecardinality of
~’. Letq(k)
be a real valuedfunction, defined
on the setof positive integers
and such thatand
Suppose
also thatfor
some constantand conszder the sum
Then we have
The constant in
Vinogradov’s symbol depends only
on the constants6, E, included, respectively,
in(2.1), (2.2)
and(4.3).
We shall prove the
Proposition
in sections 5 - 8. In this section we shalluse it to establish the Theorem.
Let ,~’ be the set of
integers n
E(N/2, N] satisfying
n - 4(mod 24)
andwhich cannot be
represented
in the form(1.4)
withprimes
Pi, P2, P3 and with r =P11.
Let F be thecardinality
of .~. We shall establish thatObviously,
thisimplies
the estimate( 1.5) .
To
study
theequation (1.4)
with analmost-prime
variable x weapply
aweighted
sieve of Richert’stype.
Let q, v,
vl, 9
be constants such thatDenote
and
Consider the sum
where
It is clear that
where
T1
is the contribution of the terms for which > 0. We decom- posef1
as(4.13)
where
F2
is the contribution of the terms with x == 0(mod p2)
for someprime
p E[z, zl)
and whereF3
comes from the other terms.We note that the congruence condition n - 4
(mod 24)
and the sizeconditions on pi in the domain of summation in
(4.10) imply
that thereare no terms with
(6, x)
> 1 counted in Tand, respectively,
inr3.
Hencethe condition
(~, P)
= 1 from the domain of summation in(4.10)
can bereplaced by (x. 6P)
= 1.Furthermore,
if r issquarefree
with respect to theprime
numbers p E[z, zi),
if(~, 6P)
= 1 andA(r)
>0,
thenFor
explanation
we refer the reader to Halberstam and Richert[4], chap-
ter
9, p.256.
From this
point
onwards we assume also thatThen
using (4.14), (4.15)
and the definition of the set 17 we conclude that the sumF3
isempty,
i.e.Indeed,
if this were not true, then for some n E .~ theequation (1.4)
wouldhave a solution P1,P2,P3, x with x
satisfying (4.14). However,
this is notpossible
due to(4.15)
and the definition of T.It is easy to estimate
F2
from above. We haveand, having
in mind(4.7)
and(4.8),
weget
From
(4.12), (4.13), (4.16)
and(4.17)
it follows thatWe shall now estimate T from below.
Using (4.10)
and(4.11)
we repre-sent it in the form
where
and where
F5
comes from the second term in theright-hand
side of(4.11).
Changing
the order of summation weget
To find a non-trivial lower bound for r we have to estimate
r4
from below andr5
from above. We shallapply
Rosser’s sieve(see
Iwaniec[12], [13]).
First we
get
rid of the condition(x, P)
= 1 from the domains of summa-tion in
(4.20)
and(4.21) by introducing
theweight
Denote
by ~-(d)
the lower Rosser function of order D and for eachprime
p E
[z, zl)
denoteby A) (d)
the upper Rosser function of orderD/p. They satisfy
and
Furthermore,
letf (s)
andF(s)
be the functions of the linear sieve. We considerseparately
the cases5 I nand 5 f n
and use(3.15) - (3.17), (4.7) -
(4.9)
to find thatwhere
For the definition and
properties
of Rosser’s functions and the functionsf (s), F(s)
as well as forexplanation
of(4.23) - (4.27)
we refer the readerto Iwaniec
[12], ~13~.
Consider the sum
F4.
From(2.5), (4.20), (4.22), (4.23)
and(4.25)
we getwhere
and where
Consider now
F5 - Using (4.21), (4.22)
and(4.25)
we find thatwhere
r7
comes from thequantity
from theright-hand
side of(4.25).
Changing
the order of summation andapplying (2.5), (4.9)
and(4.24)
wewrite
IF7
in the formwhere
Using (4.8), (4.9), (4.23), (4.24), (4.30)
and(4.32)
we see that both func-tions
q’(k)
andq"(k) satisfy
the conditions(4.1)
and(4.2).
We consider the sumsand
using
theProposition
we conclude thatAccording
to(4.19), (4.29), (4.31)
and(4.34)
we haveConsider more
precisely
the sumsF8
andr9.
From(2.6), (3.10), (3.18), (4.30)
and(4.33)
it follows thatand, respectively,
weapply (2.6), (3.10), (3.18)
and(4.33)
toget
Furthermore, according
to(3.12), (4.24)
and(4.32)
we haveFrom
(2.8), (3.14) - (3.17), (4.26) - (4.28)
and(4.36) - (4.38)
we obtainwhere
Now we
apply
thearguments
of Halberstam and Richert~4~, chapter 9,
p. 246 and we find
where
We
specify
the constants includedby
It is easy to see that
they satisfy
the conditions(4.7)
and(4.15).
Further-more,
using
Lemma 9.1 of Halberstam and Richert[4],
we canverify
thatif q, v, vl and 0 are
specified by (4.41),
thenFrom
(2.8), (3.14), (4.28), (4.39), (4.40)
and(4.42)
weget
where Co > 0 is a constant.
Inequalities (4.18), (4.35)
and(4.43) imply
We are now in a
position
toapply
the main idea ofWooley ~22~.
SinceE > 1 we can omit the second term from the
right-hand
side of(4.44).
Weget
which
implies
the estimate(4.6)
and proves the Theorem.5. The
proof
of theProposition
-beginning
We represent the sum
I(n, k;),
definedby (2.5),
in the formwhere
The
integration
in(5.1)
can be taken over any interval oflength
oneand,
in
particular,
over, _ _ "1 - - " .
We
represent Jo
as an union ofdisjoint Farey
intervalswhere
and where
q’
andq"
arespecified by
the conditions(5.4)
Xq + q’, q + q" aq’ ==
1(mod q) , aq" - -1 (mod q) (for
more details we refer thereader,
forexample,
toHardy
andWright [5], chapter 7).
Consider the set
where
It is clear that if
1 a q P, (a, q)
=1,
then91(q, a)
cB(q, a),
hencewe can represent
Jo
in the formwhere
Hence we have
where
From
(4.4), (5.10)
and(5.11)
we see that the sum E can berepresented
as
where
To prove the
Proposition
we have to estimate the sums£1 and £2.
We shallconsider
£1
in section 6 and£2 -
in sections 7 and 8.6. The estimation of the sum
~l
We
apply
Lemma 3.1of Wooley [22],
which is based on the earlier result ofBauer,
Liu and Zhan[1].
It states that if the set .M is definedby (5.5),
then for any
integer h
such that 1 hN,
we havewhere
Using (5.2)
and(5.11)
we write7i
in the formWe now
apply (2.3), (2.4)
and(6.1)
toget
11B I t,,,
where
We note
that,
due to the choice of ourweight w(x),
it is not necessary to consideranalogs
of formula(6.1)
fornon-positive integers h,
as it was donein
[22].
Using (4.2), (5.13)
and(6.3)
we findwhere
Consider first
£i2).
* It was established in[22]
thatso,
using (2.1), (6.5), (6.7)
and(6.8)
we findwhere
It is clear that
Applying
Theorem 3 of Hua[11]
we findfor some absolute constant B > 0. From
(6.9)
and(6.10)
we conclude thatif E > B +
1,
which we shall assume, thenConsider now the sum
£i1).
First weapply (6.2)
and(6.4)
to write theexpression l(l)
in the formwhere
Furthermore,
we havewhere
Obviously,
if(k, q) t
m, then W* = 0. If(k, q) I
m, then there existsunique h
=(mod [k, q]),
such that thesystem
of congruences in the domain of summation of(6.14)
isequivalent
to r m h(mod [k, q]).
Weapply
Poisson’s summation formula and use(2.3)
and(2.4).
After somecalculations we
get
1 then we
integrate
the lastintegral by
parts m times.Having
inmind the conditions k
D, q
P andusing (2.1)
and(4.3)
we find thatthe
integral
iswhere the constant in
Vinogradov’s symbol depends only
on m. From thisobservation we conclude that the contribution to
W*, coming
from theterms
1,
isnegligible.
Moreprecisely,
we havewhere
K(n)
is definedby (2.7)
and where the constant A > 0 isarbitrarily large.
From
(2.10), (2.11), (6.12), (6.13)
and(6.15)
weget
It remains to take into account also
(2.6), (2.8), (2.9)
and(3.9)
and we findThis formula and
(2.1), (4.2), (6.7) imply
We note that any,
arbitrarily
smallpositive
value of the constant 6 from the definition ofP,
suffices for theproof
of the last estimate. Thishappens
wherever P occurs, so any progress in
obtaining asymptotic
formulas oftype (6.1)
forlarger
sets ofmajor
arcs is not relevant to ourproblem.
Finally,
from(6.6), (6.11)
and(6.16)
we obtain the estimate7. The estimation of the sum
62
Using (5.11)
and(5.13)
werepresent
the sum92
in the formwhere
Applying
H61der’sinequality
weget
where
For
T1
weapply
the well known estimate,,1
Consider
Tz.
From(7.1)
and(7.3)
weget
where
To estimate the first term from the last line of
(7.5)
weapply
the in-equality
Its
proof
is very similar to theproof
of formula(4.3)
from author’s pa- per[20],
so we omit it.In the next section we shall estimate
~(l)
and we shall prove thatFrom
(2.1), (4.3), (7.5), (7.7)
and(7.8)
we obtainApplying (7.2), (7.4)
and(7.9)
weget
It remains to combine
(5.12), (6.17)
and(7.10)
and we find that the estimate(4.5) holds,
which proves theProposition.
8. The estimation of
For any
integers
a, q,satisfying 1 a q X, (a, q)
=1,
we considerthe set
9R(q, a),
definedby
where the
integers q’
andq"
arespecified by (5.4). Using (5.3), (5.6), (5.8), (5.9), (7.6)
and(8.1)
we find thatX,
then we havewhere
Formula
(8.3)
is aspecial
case of Lemma 12 from the paper[8] by
Heath-Brown and the author.
From
(7.1)
and(8.3)
it follows that theintegrand
in(8.2)
can be repre-sented in the form
where
and where n, k and u are four dimensional vectors with
components,
re-spectively,
ni,ki
and ui, i =1, 2, 3, 4.
We substitute the
expression (8.6)
for theintegrand
in(8.2). Then,
inorder to
apply
Kloosterman’smethod,
wechange
the order ofintegration
and summation over a. We find that
where
To
proceed
further we have to express the conditiona) 3 /3
in moreconvenient form. We do this in a standard way
by introducing
a functionQ(v,
q,~3),
defined forintegers
q, v such that 1 qX, -q/2
vq/2
and real numbers
{3
E[_(qX)-l, (qX)-l].
For fixed v and q this function isintegrable
with respect to/3
andsatisfy
Furthermore,
if(a, q)
=1,
if a(mod q)
is definedby
aa - 1(mod q)
andif q’ and q"
arespecified by (5.4),
thenA construction of a function with these
properties
isavailable,
forexample,
in Heath-Brown
[7],
section 3.Using (8.13)
we can express the conditiona) 3 j3
from the domainof summation in
(8.10).
If P
q X, then, according
to(8.1), (8.11)
and(8.13),
we find thatN
If q
P then weintegrate
overhence, having
in mind(8.1),
we see that the conditionOO1(q, a) 3 /3
in(8.10)
is
equivalent
to(
2013., ..] 1
30.
Therefore we can useagain (8-13)
is
equivalent {3.
Therefore we can useagain (8.13)
to obtain
(8.14).
We conclude that
~(l)
can be written in the formwhere
Using (4.1), (4.2), (8.12)
and(8.15)
we find that, ,- ,
where
E’
means that the summation is taken oversquarefree odd ki
andConsider the sum W. From
(8.4), (8.8)
and(8.16)
we see that it can bewritten as
In this form the sum W is very similar to a sum, considered
by
Brfdernand
Fouvry [2]. Applying
themethod, developed
for theproof
of Lemma 1from
[2],
we find that if~2
aresquarefree
oddintegers,
thenConsider the
quantity T,
definedby (8.18). Obviously
Applying (8.5), (8.7), (8.9), (8.21)
and Lemma10(ii)
of[8]
weget
If n = 0 and P
q X,
then we use(8.5), (8.9), (8.21)
andapply
thewell known estimate
We
get
If n = 0 and q
P,
then we use(2.2), (8.11), (8.18)
and(8.23)
to obtainFrom
(8.17)
we find thatwhere is the contribution of the terms from the
right-hand
side of(8.17)
such that
n ~ 0, ’b2
comes from the terms with n = 0 and Pq ~
Xand, finally, ~3
comes from the terms for which n = 0and q
P.To estimate we use
(8.20)
and(8.22)
and weget
After some standard
calculations,
which are very similar to those in sec-tion 5 of
[8],
we obtainWe leave the verification of this estimate to the reader.
For
~2
weapply (8.20)
and(8.24)
toget
To estimate
Q3
weapply, respectively, (8.20)
and(8.25)
and we find, - , ’I. , - , .
The estimate
(7.8)
is a consequence of(8.26) - (8.29).
This proves the
Proposition
and now theproof
of the Theorem is com-plete.
Added in
proof:
Twointeresting results, concerning
thequantity E1 (Y),
defined in section
1, appeared
since the present paper was submitted forpublication. J.Liu, Wooley
and Yu[17]
established the estimateE1 (Y)
«Y’18+6
and veryrecently
Harman and Kumchev[6] proved
thatEl (Y)
GCReferences
[1] C. BAUER. M.-C. LIU. T. ZHAN, On a sum of three prime squares. J. Number Theory 85 (2000), 336-359.
[2] J. BRÜDERN. E. FOUVRY, Lagrange’s Four Squares Theorem with almost prime variables.
J. Reine Angew. Math. 454 (1994), 59-96.
[3] G. GREAVES, On the representation of a number in the form x2 + y2 + p2 + q2 where p and q are odd primes. Acta Arith. 29 (1976), 257-274.
[4] H. HALBERSTAM. H.-E. RICHERT, Sieve methods. Academic Press, 1974.
[5] G. H. HARDY. E. M. WRIGHT, An introduction to the theory of numbers. Fifth ed., Oxford
Univ. Press, 1979.
[6] G. HARMAN, A. V. KUMCHEV, On sums of squares of primes. Math. Proc. Cambridge Philos.
Soc., to appear.
[7] D.R. HEATH-BROWN, Cubic forms in ten variables. Proc. London Math. Soc. 47 (1983),
225-257.
[8] D.R. HEATH-BROWN. D.I.TOLEV, Lagrange’s four squares theorem with one prime and three almost-prime variables. J. Reine Angew. Math. 558 (2003), 159-224.
[9] L.K. HUA, Some results in the additive prime number theory. Quart. J. Math. Oxford 9 (1938), 68-80.
[10] L.K. HUA, Introduction to number theory. Springer, 1982.
[11] L.K. HuA, Additive theory of prime numbers. American Mathematical Society, Providence, 1965.
[12] H. IWANIEC, Rosser’s sieve. Acta Arith. 36 (1980), 171-202.
[13] H. IWANIEC, A new form of the error term in the linear sieve. Acta Arith. 37 (1980),
307-320.
[14] H.D. KLOOSTERMAN, On the representation of numbers in the form ax2 + by2 + cz2 + dt2.
Acta Math. 49 (1926), 407-464.
[15] J. LIU, On Lagrange’s theorem with prime variables. Quart. J. Math. Oxford, 54 (2003),
453-462.
[16] J. LIU, M.-C. LIU, The exceptional set in the four prime squares problem. Illinois J. Math.
44 (2000), 272-293.
[17] J.LIU, T. D. WOOLEY. G. Yu, The quadratic Waring-Goldbach problem. J. Number Theory,
107 (2004), 298-321.
[18] V.A. PLAKSIN, An asymptotic formula for the number of solutions of a nonlinear equation for prime numbers. Math. USSR Izv. 18 (1982), 275-348.
[19] P. SHIELDS, Some applications of the sieve methods in number theory. Thesis, University of Wales 1979.
[20] D.I. TOLEV, Additive problems with prime numbers of special type. Acta Arith. 96, (2000),
53-88.
[21] D.I. TOLEV, Lagrange’s four squares theorem with variables of special type. Proceedings of
the Session in analytic number theory and Diophantine equations, Bonner Math. Schriften, Bonn, 360 (2003).
[22] T.D. WOOLEY, Slim exceptional sets for sums of four squares, Proc. London Math. Soc.
(3), 85 (2002), 1-21.
Doychin TOLEV
Department of Mathematics Plovdiv University "P. Hilendarski"
24 "Tsar Asen" str.
Plovdiv 4000, Bulgaria
E-mail : dtolev@pu.acad.bg